AI Paper Atlas

Research map / Video / Video segmentation and tracking

MindCross: Fast New Subject Adaptation with Limited Data for Cross-subject Video Reconstruction from Brain Signals

Xuan-Hao Liu, Yan-Kai Liu, Tianyi Zhou, Bao-Liang Lu, Wei-Long Zheng

AAAI 2026 · Video › Video segmentation and tracking · approach: Sign language and physiological video analysis · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add reconstruction, domain alignment and domain classification losses to the alignment-only loss (ablation)MindCross · SEED-DV (EEG)video-based 2-way accuracy: 0.756 -> 0.789Improved
Add reconstruction, domain alignment and domain classification losses (ablation)MindCross · SEED-DV (EEG)frame-based 2-way accuracy: 0.693 -> 0.757Improved
Add reconstruction, domain alignment and domain classification losses (ablation)MindCross · CC2017 (fMRI)video-based 2-way accuracy: 0.812 -> 0.829Improved
Add difference loss L_diff in training (ablation)MindCross · SEED-DV (EEG)video-based 2-way accuracy: 0.789 -> 0.786No clear effect
Add difference loss L_diff in training (ablation)MindCross · CC2017 (fMRI)frame-based 2-way accuracy: 0.810 -> 0.802No clear effect
Add Top-K collaboration module at test time (ablation)MindCross · SEED-DV and CC201740/50-way video, frame, PSNRImproved
Use K=2 instead of K=1 in Top-K collaboration (ablation)MindCross · SEED-DV and CC2017decoding metricsNo clear effect
Pretrain on existing subjects then calibrate instead of training on the new subject from scratch (ablation)MindCross · new-subject adaptation (limited data)decoding metricsImproved
Calibrate with L_align plus reconstruction and difference losses instead of L_align only (ablation)MindCross · new-subject adaptationdecoding metricsImproved
Calibrate only the new subject's specific encoder instead of fine-tuning (MindBridge)MindCross · SEED-DV new subject, 200 samplescalibration time (s): 5.104 -> 1.090Improved
Calibrate only the new subject's specific encoder instead of fine-tuning (GLFA)MindCross · CC2017 new subject, 500 samplescalibration time (s): 17.513 -> 2.724Improved
Use a single cross-subject MindCross model instead of per-subject modelsMindCross · SEED-DVvideo-based 2-way accuracy: 0.800 (EEG2Video, 20 models) vs 0.786Mixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy